Image-based Level Measurement in Flotation Cells using Particle Filters

This paper is concerned with the development and implementation of an image-based sensor for the detection and control of the Bitumen froth and middlings interface in a separation or flotation cell for the extraction of bitumen. Images obtained from a camera are transformed through an edge detection algorithm and combined with a Particle filter algorithm to estimate the interface. Particle filters are a general class of Bayesian filters, of which the Kalman filter is a special case when the system is linear and the noise processes are Gaussian. This paper provides a tutorial introduction to Particle filters and then considers the application of such a filter to estimate the Bitumen froth and middlings interface from images captured from the sight glass of the separation cell. Industrial results show that the interface level detection algorithm is robust to lighting changes. They also show highly improved control performance when these interface level estimates are used in feedback control.